Container Throughput Prediction Based On A Novel Hybrid Model SSA-ICEEMD-CS-LSSVR

At present,China is at a critical period of “The Belt and Road” initiative,and the construction of ports plays a decisive role in promoting the “21st-Century Maritime Silk Road”.However,accurately analyzing and predicting the ports' container throughput have great significance on formulating relevant policies for ports construction operation.This paper takes the container throughput data of Guangzhou port and Tianjin port as an example and build a novel hybrid ensemble model for container throughput prediction.In this paper,Singular Spectrum Analysis(SSA)is firstly employed to extract tendency of the original series and then get the reconstructed series.Secondly,Improved Ensemble Empirical Mode Decomposition(ICEEMD)algorithm is used to decompose the reconstructed sequence into several IMFs and one residual series.Thirdly,Least Square Support Vector Regression(LSSVR)models which parameters c and g are optimized by Cuckoo Search(CS)algorithm are used for training prediction models.Finally,the proposed model SSA-ICEEMD-CS-LSSVR is applied to the container throughput history data of Guangzhou port and Tianjin port.The empirical research results show that the proposed model has higher prediction accuracy than other benchmark models.